T5 Qg Webnlg Synth En

🚀 t5-qg_webnlg_synth-en

This is a Data Question Generation model. It takes a structured table and a conditioned answer as input, then generates relevant questions. It's also a part of the QuestEval metric but can be used independently for question generation.

🚀 Quick Start

✨ Features

• Based on the T5-small architecture.
• Generates questions from a structured table and a given answer.
• Can be used independently or as part of the QuestEval metric.

📦 Installation

There is no specific installation command provided in the original document. However, the code to load the model is as follows:

from transformers import T5Tokenizer, T5ForConditionalGeneration

tokenizer = T5Tokenizer.from_pretrained("ThomasNLG/t5-qg_webnlg_synth-en")

model = T5ForConditionalGeneration.from_pretrained("ThomasNLG/t5-qg_webnlg_synth-en")

💻 Usage Examples

Basic Usage

from transformers import T5Tokenizer, T5ForConditionalGeneration

tokenizer = T5Tokenizer.from_pretrained("ThomasNLG/t5-qg_webnlg_synth-en")

model = T5ForConditionalGeneration.from_pretrained("ThomasNLG/t5-qg_webnlg_synth-en")

Advanced Usage

You can play with the model using the inference API. The text input format should follow this template (accordingly to the training stage of the model):

text_input = "{ANSWER} </s> {CONTEXT}"

where CONTEXT is a structured table that is linearised this way:

CONTEXT = "name [ The Eagle ] , eatType [ coffee shop ] , food [ French ] , priceRange [ £ 2 0 - 2 5 ]"

📚 Documentation

Model description

This model is a Data Question Generation model based on T5-small, that generates questions, given a structured table as input and the conditioned answer. It is actually a component of QuestEval metric but can be used independently as it is, for QG only.

Training data

The model was trained on synthetic data as described in Data-QuestEval: A Referenceless Metric for Data to Text Semantic Evaluation.

Citation info

@article{rebuffel2021data,
  title={Data-QuestEval: A Referenceless Metric for Data to Text Semantic Evaluation},
  author={Rebuffel, Cl{\'e}ment and Scialom, Thomas and Soulier, Laure and Piwowarski, Benjamin and Lamprier, Sylvain and Staiano, Jacopo and Scoutheeten, Geoffrey and Gallinari, Patrick},
  journal={arXiv preprint arXiv:2104.07555},
  year={2021}
}

📄 License

This model is released under the MIT license.